Prior knowledge of the noise present in a color image acquisition device is very important for the recovery of a spectral reflectance of an object being imaged, since the recovery performance is greatly influenced by the noise.In the previous paper (IEEE Trans. Image Process. 15, 1848 (2006)), the author proposed a new model to estimate the noise variance of an image acquisition system by assuming that the noise variance in each channel is equal and showed that this model is very useful to accurately recover a spectral reflectance of an imaged object. This paper describes an extended model for the estimation of the covariance matrix of the noise present in an image acquisition system without the assumption. It is demonstrated that the proposal overfits the noise covariance matrix to learning samples and that the recovery performance for the test samples is poor compared with the previous model. However this overfitting means that the estimates are correctly performed using the proposed model. The new model is effective in analyzing the noise present in an image acquisition system.